AlphaFold adds 8000 viral protein dimers from 23 families
Researchers added more than 8000 virus protein dimers to the AlphaFold database, from 23 virus families with human-infecting members. The team judged 2749 homodimers and 5279 heterodimers accurate enough to include. EMBL reported the same day's release of predicted complexes for more than 2800 viruses, with priority for families known to infect humans.
Science··Night
8000 dimers, 23 families
Researchers today added more than 8000 virus protein dimers to the AlphaFold Protein Structure Database. The entries come from 23 virus families, each with human-infecting members. The team judged 2749 homodimers and 5279 heterodimers accurate enough to include. The predictions still await experimental checks. EMBL reported on 24 September that predicted protein-complex structures for more than 2800 viruses are now in the AlphaFold Database. The set gives priority to viral families known to infect humans. The page does not give the dimer counts.[1], [2]
The dimers cover more than 2,800 viruses
Nature's count is dimers: more than 8000, split as 2749 homodimers and 5279 heterodimers, drawn from 23 families. EMBL's count is viruses: more than 2800, with priority for families known to infect humans. Both pages date the release to 24 September. A reader who adds the two figures together is mixing a dimer total with a virus total. Joe Grove at the University of Glasgow describes many viral proteins as acting with partners rather than alone. That sentence is on the Nature account. The homodimer and heterodimer figures stay on the Nature account. The experimental-check limit is also Nature's: the predictions still await experimental checks.[1], [2]
The predictions still await experimental checks
Nature stops at the inclusion cut: 2749 homodimers and 5279 heterodimers were accurate enough to include, and the predictions still await experimental checks. It does not list every family by name in the summary used here. The 8000 figure is a dimer total, not a count of patients or of outbreaks. The reader sees 8000 dimers, 23 families, 2749 homodimers and 5279 heterodimers in this account. The predictions still await experimental checks. Joe Grove describes many viral proteins as acting with partners. The dimer total is not a patient count. Each family has human-infecting members. The cut ends here. The account stops at that inclusion cut and does not name every family. EMBL stops at the virus total, more than 2800, and at the priority for families known to infect humans. It does not repeat 2749 or 5279. The shared fact is the 24 September AlphaFold release aimed at human-infecting virus families. The two integers answer two questions.[1], [2]
